The genomic profiling and MAMLD1 expression in human and canines with Cushing’s disease

Background Cushing’s disease (CD) is defined as hypercortisolemia caused by adrenocorticotropic hormone (ACTH)-secreting pituitary adenomas (corticotroph PA) that afflicts humans and dogs. In order to map common aberrant genomic features of CD between humans and dogs, we performed genomic sequencing and immunostaining on corticotroph PA. Methods For inclusion, humans and dog were diagnosed with CD. Whole exome sequencing (WES) was conducted on 6 human corticotroph PA. Transcriptome RNA-Seq was performed on 6 human and 7 dog corticotroph PA. Immunohistochemistry (IHC) was complete on 31 human corticotroph PA. Corticotroph PA were compared with normal tissue and between species analysis were also performed. Results Eight genes (MAMLD1, MNX1, RASEF, TBX19, BIRC5, TK1, GLDC, FAM131B) were significantly (P < 0.05) overexpressed across human and canine corticotroph PA. IHC revealed MAMLD1 to be positively (3+) expressed in the nucleus of ACTH-secreting tumor cells of human corticotroph PA (22/31, 70.9%), but absent in healthy human pituitary glands. Conclusions In this small exploratory cohort, we provide the first preliminary insights into profiling the genomic characterizations of human and dog corticotroph PA with respect to MAMLD1 overexpression, a finding of potential direct impact to CD microadenoma diagnosis. Our study also offers a rationale for potential use of the canine model in development of precision therapeutics. Supplementary Information The online version contains supplementary material available at 10.1186/s12902-021-00845-z.


Background
Corticotrope pituitary adenomas (PA) are benign tumors of the anterior pituitary gland that secrete excessive amounts of adrenocorticotropin hormone (ACTH), causing Cushing's disease (CD) [1]. The precursor polyprotein to ACTH, proopiomelanocortin (POMC), is synthesized within residential anterior pituitary corticotropic cells [2]. When ACTH is secreted, zona fasciculata cells within the adrenal cortex release cortisol, which exerts diverse physiological systemic effects [3,4]. Mortality is increased up to 5 times in CD compared with the general population, but significantly improves after achievement of normocortisolemia by treatment [5,6].
CD treatment options are diverse. Surgery is the first choice treatment and removal of the tumor results in remission in 75-90% of patients with a 35% biochemical recurrence [4,[7][8][9][10]. The remainder have either small tumors not identified at surgery or nonresectable tumors that invade the cavernous sinus [11]. Other options include radiotherapy and bilateral adrenalectomy, however these procedures have risks of permanent cortisol and aldosterone deficiency along with potentiation of aggressive tumor growth and inevasible behavior (Corticotroph Tumor Progression) [7]. Finally, while medical treatment options have expanded to include somatostatin receptor ligands, cortisol synthesis inhibitors, glucocorticoid receptor blockers, and dopamine agonists, long-term safety data continues to be lacking [12][13][14][15]. Taken together, many of the available compounds target the adrenal gland or suppress ACTH secretion, but not the source of the secreting PA, therefore there is a strong need for further research on this topic.
Dogs may be viable models for targeted interventions in CD. Dogs share an 84% genetic overlap with humans, and they naturally develop Cushing's disease from the intermediate pituitary lobe with an incidence of 1500: 1,000,000 and recurrence after hypophysectomy of 27%, analogous to humans [28][29][30][31][32]. A canine disease model has advantages over transgenic mouse models, particularly since dogs share similar environmental exposures and stressors to humans [29]. However, while TBX19 expression were demonstrated in dog PA, USP8 mutations were not identified as frequently mutant [33]. This prompted us to investigate a more complete catalogue of genomic aberrations that may lead to dysfunctional protein expression observed in CD and could potentially lead to the development of targeted pharmaceutical interventions.
Using both human and dog corticotroph PA, our objective was to describe the common aberrant genomic features of CD between humans and dogs. We profiled genetic expression using whole transcriptome sequencing of both species. Whole exome sequencing (WES) was also performed on the human corticotroph PA and an identified target was confirmed by immunohistochemistry when compared to normal human pituitary tissue. Across experimentation, corticotroph PA were compared with normal tissue, and when possible between species analysis were also performed. Furthermore, although genetic studies have been previously conducted in humans with CD [15,18,19,24,[34][35][36][37], our results provide a broad overview of the genetic profiles between human and dog corticotroph PA.

Methods
For eligibility, human and dogs were required to have a CD diagnosis. Genomic sequencing and immunostaining were performed on corticotroph PA. Across experimentation, corticotroph PA were compared with normal tissue, and when possible between species analysis were also performed.

Human cohort
Thirty-seven CD African American patients (average 43 +/− 12 years old) with intermediate anterior sized PA (avg lateral 1.16 +/− 0.44 cm x AP 0.88 +/− 0.29 cm x cranial 0.90 +/− 0.49 cm) were recruited through the Department of Neurosurgery at Emory University Hospital. In all cases, the diagnosis of Cushing's disease entailed the following steps: 1. two or more abnormal screening tests (including lack of suppression to low-dose dexamethasone, higher than normal urinary free cortisol levels and/or higher than normal late-night salivary cortisol levels), 2. ACTH-dependent hypercortisolism (i.e. high or high-normal ACTH level) and 3. localization tests supporting autonomous pituitary ACTH secretion (cortisol suppression by more than 50% or below 5 mcg/ dL after to high-dose dexamethasone and/or ACTH stimulation after CRH test). In addition, depending on pituitary MRI findings, selected patients underwent petrosal sinus sampling which indicated a baseline centralto-periphery ACTH ratio > 2 and/or post-CRH ratio > 3). For all cases, pathology reports showed immunopositive staining for ACTH, and negative immunostaining for tumor protein p53 (TP53), follicle-stimulating hormone (FSH), thyroid stimulating hormone (TSH), luteinizing hormone (LH), growth hormone (GH), and E3 ubiquitin protein-ligase (MIB-1) proliferation index < 3%. During microsurgical transsphenoidal hypophysectomy, the PA tissue was removed with no complications.

Dog cohort
CD Dogs from canine client owners were recruited through the Department of Clinical Sciences of Companion Animals at the Utrecht University, the Netherlands. Suspicion of hypercortisolism was raised based on the characteristic clinical signs including polyphagia, polydipsia/polyuria, skin atrophy, thin hair coat, calcinosis cutis, truncal obesity (pot belly), depression, and exercise intolerance. Serum chemistry showed typical changes: alkaline phosphatase, alanine aminotransferase and total cholesterol were commonly increased in canine Cushing's disease.
Preliminary diagnosis of pituitary-dependent hypercortisolism was based on increased urinary corticoidto-creatinine ratios (UCCRs) in the first two morning urinary samples collected at home and more than 50% suppression of the UCCR in the third urine sample in the oral high dose (0.1 mg/kg) dexamethasone suppression test [32]. CD was further confirmed by measurement of elevated plasma ACTH concentration, visualization of symmetrically enlarged adrenal glands by ultrasonography, and visualization of pituitary gland enlargement with computed tomography (CT) or MRI. During microsurgical transsphenoidal hypophysectomy in client-owned dogs, the PA tissue was removed with no complications. Immediately after collection, specimens of pars distalis PA tissue were fixed in 4% neutral buffered formaldehyde, embedded in paraffin, and consecutive sections were used for histology and immunohistochemistry for ACTH, α-melanocyte-stimulating hormone (MSH), and GH [32]. Representative adenoma tissue samples were snap-frozen and stored in liquid nitrogen until analysis after histology confirmed a basophilic adenoma with ACTH immunostaining.
Pituitary specimens from 6 healthy beagle dogs and 7 client-owned dogs with CD that underwent hypophysectomy were sequenced. The cohort of client-owned dog patients included different breeds (i.e., Jack Russell Terrier, Scottish Shepherd, English Springer Spaniel, Mixed breed, French Bulldog, Chesapeake Bay Retriever, and American Staffordshire Terrier), 3 female (of which two castrated) and 4 male (of which one castrated) dogs, with a median age of 8.2 years (range 5.6-10.7 years). All owners consented to the hypophysectomy as treatment for CD. As control tissue, anterior lobes of normal pituitary glands were obtained from 6 healthy Beagle dogs euthanized in unrelated experiments (all female; age range 1.6-1.8 years).

Nucleic acid extraction and sequencing
DNA and RNA were extracted from 6 intermediate sized frozen human PA tissues using E.Z.N.A. Ⓡ kits (Omega Bio-tek, Norcross, GA). RNA was extracted from the canine healthy and tumor pituitary samples using the miR-CURY cell and plant RNA isolation kit according to the manufacturer's instruction including an on-column DNAse digestion (Qiagen, RNAse free DNAse, 79,254). Specimen quantity and quality were assessed using Qubit and Agilent Bioanalyzer.

Human exome sequence analysis
Human genomic DNA was extracted from fresh-frozen PA and normal anterior pituitary as described above. Whole human exome libraries were prepared using the Agilent SureSelect Human All Exon (version 5) per manufacturer's protocols. Libraries were paired-end sequenced at 50x coverage using and Illumina HiSeq 2000. FASTQ files were aligned to the human hg19 reference genome using BWA 0.7.5.a and duplicate reads were removed with Picard tools (Version 1.1.1) [32,38]. Deduplicated aligned (BAM) files were used for copy number estimation using Control-FREEC with recommended settings for exome sequencing [39]. Normalized FREEC output values were segmented with DNACopy [40]. Copy number assessments for 6 PA samples were matched to one normal. Human PA specimen 1, had a matched normal sample and mutations were called using Varscan2 in somatic mode under standard parameters [41]. Mutations in the other 5 samples were called using Varscan2 without a matched normal. Human PA specimen 1 was also analyzed with Mutect (version 1.1.4) using standard parameters following probabilistic indel realignment and base recalibration using GATK (version 3.3.0) [42]. GISTIC-like analysis using cghMCR were used to identify potential regions of focal gain and loss [43]. Identified single nucleotide variants (SNV) were compared to somatic variants in the COSMIC database (version 81) and gnomAD database [38,44]. All predicted mutations were annotated using ANNOVAR [45].

RNA sequence analysis
Total RNA was processed using the Illumina TruSeq RNA kit and paired-end sequenced (2 × 75) at 100 M reads per specimen on a HiSeq 2000 instrument. FASTQ files were aligned to the hg19 human or CanFam3 reference genomes using Tophat 2.0.1 using standard parameters [46]. RefSeq, CanFam3 and Ensembl transcripts were quantified using Cufflinks (version 2.0.1) [47]. Human RefSeq transcripts were also quantified using HTSeq (version 0.6.1) [48]. Mutations were called in the canine and human samples using Varscan2 as described above. Identification of differentially expressed genes (> 2 fold change) between tumor and normal tissue by simple hierarchical clustering of normalized fragments per kilobase of transcript per million mapped reads (FPKM) values were assessed separately for human and dog Cushing samples and dichotomized into over or under gene expression. With the feasible utilization of IHC for conformational protein upregulation, only shared overexpressed genes were analyzed.

Statistical analysis
Differentially expressed transcripts were converted to normalized fragments per kilobase million (FPKM) values using tools described above [47,48]. Average log transformed FPKM values for both humans and dog transcripts are expressed with standard deviations. Comparisons of each subset of genes between ACTH secreting PA and normal pituitary tissues across either the human or dog groups were considered statistically significant when P < 0.05 in a one-tail Student T-test (i.e. overexpression tail). Immunohistochemistry results were expressed as proportions, mean, standard deviations (SD) and median. Scoring agreements were expressed as a weighted Cohen's Kappa coefficient. With no censored or missing data, a sensitivity analysis were performed between stained and non-stained genes. Stata/IC (v15.0; StataCorp LP, College Station, TX, USA) statistical software was used for all analyses.

Whole exome sequencing
To identify recurrent somatic mutations and copy number abnormalities in corticotroph PA, whole-exome sequencing was performed in human PA.
Copy number analysis using the whole exome sequencing data was also performed to ascertain recurrent gains of whole chromosomes or chromosome arms (Fig. 1). Human PA specimens 2 and 5 had similar patterns with gain of 1q, 5, 7, 8q, 9, 12, 13 and 14. Specimen 1 shared gain of 7, 12 and 14 with specimens 2 and 5. Loss of 19 was shared by specimens 1, 3 and 6. No focal gains or losses enriched in PA in non-benign copy number variations were observed.

RNA-sequencing
Gene expression between Cushing's tumor and normal tissue was assessed in human and dogs.
Unsupervised hierarchical clustering showed the normal pituitary specimens clustered separately from the PA specimens for both species (Fig. 2). Heterogeneity of gene expression was found in individual PA specimens in both species, and a subset of genes that were over or under expressed relative to normal pituitary in both species organisms clustered (Fig. 2).

Immunohistochemistry
To examine the association between significantly expressed genes and protein translation, immunohistochemistry was performed on four of the eight significant genes in human PA (Fig. 5).
The strength of agreement between reviewers was almost perfect (k = 0.87). Sensitivity analysis between   stained and non-stained genes outcomes were not statistically significant (P > 0.05).

Discussion
Our WES provides the first broad overview of RNAseq gene expression profiles of PA in dog and man with CD. Mutations in USP8 as the main driver and Guanine nucleotide-binding protein G(s) subunit alpha isoforms short (GNAS) have been implicated in PA tumorigenesis [18,19,34].
While USP8 even by manual inspection of aligned DNA and RNA sequencing reads were not found, two separate non-synonymous changes were found at codon 415 of USP48 (p.M415I and p.M415V). Both have been reported as somatic variants in recent Cushing's PA and in the COSMIC database in cancers arising from the urinary tract and prostate [22]. These variants were unlikely to be either common or rare population variants as they were absent from the gnomAD database. In these variants, methionine 415 is a conserved residue falling within the Peptidase C19 ubiquitinyl hydrolase domain of USP48 and bioinformatic predictions using Polyphen 2 suggested that both changes may affected protein function.
Two specimens had Guanine nucleotide-binding protein G(q) subunit alpha q polypeptide (GNAQ) p.T96S SNVs instead of GNAS mutations. T96S has been reported in 20 COSMIC samples (COSM404628) -predominantly large intestine and skin -and may represent a third hotspot within the protein additional to the wellcharacterized Q209 and R183 positions. There have been more recent studies that have shown that no significantly recurrent mutations are identified by WES in PA [27,50]. Combined with our results, these discrepancy illustrates that PA, although rare, may also be a heterogeneous disease from the genetic perspective. Larger cohorts are needed to determine if mutations or copy number abnormalities can be used to stratify and target therapies for PA.
Along with TBX19, our preliminary findings show MAMLD1 was found to be strongly associated to PA. While a relationship between gonadotroph pituitary adenomas and MAMLD1 has been elucidated [51], little is known about the function of MAMLD1 -to the best of our knowledge this is the first report link to Cushing's disease. The Mastermind Like (MAML) family of genes vary significantly in function. MAMLD1 has a transcriptional co-activator role in regulating NOTCH signaling [39][40][41]52]. Recently, another MAML family gene, Mastermind Like Transcriptional Coactivator 3 (MAML3) has been found to be associated with hereditary paragangliomas (PGL), a type of neuroendocrine tumor that is usually derived from extra-adrenal chromaffin cells as opposed to the pituitary [53]. However, despite being in the same family, MAML3 functions as a fusion protein, whereas MAMLD1 functions as an inhibitor protein. Evidence suggests MAMLD1 may bind the recombining binding protein suppressor of hairless (RBJP) repressor protein and subsequently inhibit hairy and enhancer of split gene (HES) expression [54]. During neuronal development, HES family members play a role in corticotrophic proliferation and differentiation [55], which HES1 knockouts has lead to an underdeveloped hypermorphic pituitary [56]. Our study underscores the role MAML proteins seem to play in regulating neuroendocrine function and encourages further functional investigations.
There are several limitations to this study. First, due to the rare quantity of corticotrophic PA, only four of the eight significantly overexpressed cross-species genes in RNA-seq were able to have been randomly selected and stained for IHC from archival specimens. Although a sensitivity analysis showed no difference, continued IHC studies should be performed for non-stained genes, for MAMLD1 replication in dog PA specimens, and for exploring the value of the commonly underexpressed genes. Second, while USP8 mutations where found in 40% of human PA (Caucasian and Asian populations), similar to other studies, neither of our cohorts (2.4% chance) found USP8 mutations or statistically significant DGE of POMC (due to wider expression variance) despite staining immuno-positive for well-established CD genes (POMC, TBX19) from intermediated sized PA female samples (African American population) [18][19][20][21]. Further exploration behind different populations and PA models may be of interest to fully illustrate the genetic variations of PA. Third, appropriate tissue was not available to confirm protein expression in the dog specimens   [57,58] and no functional MAMLD1 assays were complete, our current investigation was neither a discovery cohort for direct comparison nor a functional assessment study, respectively. Rather, our cohort aimed to broadly profile the somatic aberrations of PA between dog and man.
In spite of these genetic heterogeneities, differentially expressed genes were identified that were common between human and dog PA. Our study not only highlights and contributes to the growing complex understandings of genetic mechanism of CD, but also underscores the potential clinical implications of shared mechanisms between dog and man. Since MAMLD1 appears to stain exclusively for corticotrophic tumor cells as compared to TBX19 (may stain both normal and tumor corticotrophic cells), MAMLD1 staining may be helpful in diagnosing CD, particularly in microadenomas with aberrant in symptomology and potentially recurrent when tissue quantity is limited. This may improve CD clinical outcomes and help better characterize different hormone secreting pituitary tumors. Our results also suggests that, because MAMLD1 expression appears absent in normal pituitary, targeted inhibition of MAMLD1 may be a potential potent targeted strategy for inhibiting the growth of corticotroph PA that merits further functional exploration [7]. Furthermore, dog PA may provide suitable veterinary cohorts to test novel diagnostic and therapeutic approaches that may reduce CD burden in both dog and man.

Conclusion
We highlight the first preliminary insights into profiling the genomic characterizations of human and dog corticotroph PA with respect to MAMLD1 overexpression, a finding of potential direct impact to CD microadenoma diagnosis. Our study also offers a rationale for potential use of the canine model in development of precision therapeutics.
Additional file 1: Table S1. Post-agreement Immunohistochemistry Chemistry Scoring. Individualized immunohistochemistry scoring (0 to 3+) of Cushing's pituitary adenoma human specimens from 7 antibodies (k = 0.87). Fig. S1. Difference in Magnitude of RNA-seq T-test Results. Normalized averaged FPKM values and standard errors are shown for genes suspected to be overexpressed by RNA-seq in human corticotroph pituitary adenomas (HC PA, n = 6) and dog corticotroph pituitary adenomas (DC PA, n = 7). Fig. S2. GeneCards® Summary Characteristics of Highly Expressed Genes in Humans and Dogs with Cushing's disease. Fig. S3. Theoretical Schematic Representation of the Highly Expressed Genes in Humans and Dogs with Cushing's disease.